2022-10-17 • General simulator software design
Contents
2022-10-17 • General simulator software design¶
In the previous notebook, the firing rate error in the N-to-1 simulations was fixed. We want to know re-run those simulations with actual lognormal Poisson inputs.
When writing the network simulation code, the N-to-1 simulation code was copied and adapted. I.e. there is duplication in functionality, and divergence in their APIs. It’s time for consolidation. Advantage: easier to also investigate LIF/EIF neurons, different neuron types, etc.
Differential equations¶
izh = @eqs begin
dv/dt = (k*(v-vₗ)*(v-vₜ) - u - I_syn) / C
du/dt = a*(b*(v-vᵣ) - u)
I_syn = gₑ*(v-Eₑ) + gᵢ*(v-Eᵢ)
dgₑ/dt = -gₑ / τ # Represents sum over all exc synapses
dgᵢ/dt = -gᵢ / τ
end;
izh
SpikeLab.Model_
with variables {v, u, I_syn, gₑ, gᵢ}
and parameters {C, Eᵢ, Eₑ, a, b, k, vᵣ, vₗ, vₜ, τ}
izh.generated_func
:((diff, vars, params)->begin
@unpack (v, u, I_syn, gₑ, gᵢ) = vars
@unpack (C, Eᵢ, Eₑ, a, b, k, vᵣ, vₗ, vₜ, τ) = params
diff.v = ((k * (v - vₗ) * (v - vₜ) - u) - I_syn) / C
diff.u = a * (b * (v - vᵣ) - u)
vars.I_syn = gₑ * (v - Eₑ) + gᵢ * (v - Eᵢ)
diff.gₑ = -gₑ / τ
diff.gᵢ = -gᵢ / τ
end)
Initialize buffers¶
params = (
# Cortical regular spiking (same as always)
C = 100 * pF,
k = 0.7 * (nS/mV),
vₗ = - 60 * mV,
vₜ = - 40 * mV,
a = 0.03 / ms,
b = - 2 * nS,
# Not in model eqs above (yet)
vₛ = 35 * mV, # spike
vᵣ = - 50 * mV, # reset
Δu = 100 * pA,
# Synapses
Eₑ = 0 * mV,
Eᵢ = -80 * mV, # Higher than Nto1 (was -65); same as nets.
τ = 7 * ms,
)
init = (
v = params.vᵣ,
u = 0 * pA,
gₑ = 0 * nS,
gᵢ = 0 * nS,
I_syn = 0 * nA,
);
vars = CVec{Float64}(init)
diff = similar(vars)
diff .= 0
ComponentVector{Float64}(v = 0, u = 0, gₑ = 0, gᵢ = 0, I_syn = 0)
duration = 1 * second
Δt = 0.1 * ms
Nt = to_timesteps(duration, Δt)
10000
for ti in 1:Nt
izh.f(diff, vars, params)
vars .+= diff .* Δt
end
Revise.revise(SpikeLab);
Revise.revise(VoltoMapSim);
poisson_spikes(5Hz, 20minutes)
6086-element Vector{Float64}:
0.0584
0.246
0.264
0.304
0.318
0.369
0.445
0.821
0.9
1.04
1.28
1.31
1.35
⋮
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
1.2E+03
(julia.exe:17984): Gdk-CRITICAL **: 13:16:50.068: gdk_seat_default_remove_tool: assertion 'tool != NULL' failed
(julia.exe:17984): Gdk-CRITICAL **: 13:16:50.284: gdk_seat_default_remove_tool: assertion 'tool != NULL' failed
(julia.exe:17984): Gdk-CRITICAL **: 13:16:50.420: gdk_seat_default_remove_tool: assertion 'tool != NULL' failed
?poisson_spikes
search: poisson_spikes gen_Poisson_spikes
poisson_spikes(r, T)Generate a Poisson spike train with firing rate $r$ on the time interval $[0, T]$.
More precisely, simulate a Poisson process by drawing inter-spike-intervals from an Exponential distribution with rate parameter $r$, accumulating them until $T$ is reached. The number of spikes $N$ in $[0, T]$ will be Poisson-distributed, with mean = $rT$.
The output is a length-$N$ (i.e. variable length) vector of spike times.
Pause for now, and continue with the science.